vgcwfandomcom-20200215-history
Talk:VGCW Computer Power Rankings
As a baseball fan, I like the idea of this, but if we're going to have an "objective" ranking system that goes beyond simple win% I really do want to know what goes into these numbers, and I suspect others will too. If you could explain how you arrived at these numbers that would be great. I can sort of get the picture from your strikethrough section, but it's not completely clear. The Awesomest (talk) 17:05, May 18, 2013 (UTC) Easily done! Talk page is probably the best place for this anyway. So, it's a modified version of the Elo rating system that is used in Chess, with a base/starting ranking of 1000, a K-factor of 200, and an expectation scaling factor (I forgot the technical term for it) of 500. When a match is played, it calculates the expected win probability of each wrestler, based on the formula: Win% = 1 / (1 + 10 ^ ( (rating_opponent - rating_self) / 500 ) That 500 is the expectation scaling factor mentioned above. So, if someone who is at a rating of 1000 plays someone with a 1500 rating, their win % is 1 / (1 + 10 ^ ( (1500 - 1000) / 500) = 1 / (1 + 10) = 1/11, or 0.091. You can do the same thing with all pairs of ratings, if they're the same then both have a 0.5 chance of winning. One person wins and one person doesn't. The one that wins, has an Actual Win% of 1, the one that doesn't, 0. Each wrestler's rating is updated with the formula New Rating = Old Rating + 200 * (Actual Win% - Expected Win%) In the example above, if the person who was 500 points worse going into the match won, his new rating would be 1000 + 200 * (1 - 0.091) = 1181.8 That's the core of the system. I give extra weight to #1 contender matches (150% as much of a swing, so effectively their K is 300), to title matches (200% as much, K~400) and to more-than-2 wrestler fights (also 150%). Matches with more than two wrestlers are a bit tetchy, but how I'm handling them here is by treating them as a match with the average opponent's ranking; that is, if a 1000, 1200, 1100, and 1300-rated wrestler are in a four way, it's treated as though the 1000-rated wrestler is fighting someone with an (average of 1100, 1200, 1300 = 1200) rated wrestler, the 1300-rated wrestler is fighting someone with a 1100 rating, and so on. Anything else you'd like to know? No secrets in here, that would be silly. Mukaikubo (talk) 17:28, May 18, 2013 (UTC) This is fantastic stuff! Perhaps we can covert these numbers to moneyline odds (for all those who are sportsbettors like me) and set up a vBookie system so we can bet our hard earned Bison dollars on some of these matches. Uknowwhatitis (talk) 19:59, May 18, 2013 (UTC) Here's another question: you say you don't count handicap matches. I assume you mean handicap matches that end in losses? In the event that, say, Barret had won over The Practice, how would you have ruled that? Django Boigas (talk) 20:08, May 19, 2013 (UTC) I don't have an account on the SA forums, but based on the % to win posted there, here would be the sportsbetting moneyline odds for both matchups in the "GurlGamer" #1 contenders tournament (if there was a sports bookie setting odds and stuff): 'Terra with a 68% chance to win: ' Terra -213/Jessie+203 (in sportsbetting talk, -213 means you are betting $2.13 to win $1; for Jessie, +203 means you are betting $1 to win $2.03). 'Daisy with a 55% chance to win: ' Daisy -122/Morrigan +112 (For Daisy, -122 means you are betting $1.22 to win $1; for Morrigan, +112 means you are betting $1 to win $1.12). This is for entertainment purposes ONLY!!! Seriously!!! :P Uknowwhatitis (talk) 01:36, May 20, 2013 (UTC)